Impacts of the Mobile Internet on Transportation Cyberphysical Systems:
Traffic Monitoring using Smartphones
Daniel B. Work and Alexandre M. Bayen
Abstract— This article describes how the mobile internet is themselves to remain relevant as the internet goes mobile.
changing the face of the transportation cyberphysical system In the context of traffic monitoring, the examples below are
at a rapid pace. In the last five years, cellular phone tech- eloquent and show the importance of information technology
nology has leapfrogged several attempts to construct dedicated
infrastructure systems to monitor traffic. Today, GPS equipped for transportation systems. In late 2007, Google made a move
smartphones are progressively morphing into an ubiquitous towards the phone industry with the launch of the Open
traffic monitoring system, with the potential to provide traffic Handset Alliance and the Linux-based Android platform
information in real time for the entire transportation network. (leading to the T-Mobile G1 Google phone). In part because
Traffic information systems are one of the first instantiations of of the pressure to use open platforms enhanced by the Google
the potential of participatory sensing for large scale cyberphys-
ical infrastructure systems. While mobile device technology is OS, Nokia, who manufactures 40% of the cell phones in
very promising, fundamental challenges remain to be solved, in the world, purchased Symbian, which licenses the operating
particular in the fields of modelling and data assimilation. system running on more than half of the smartphones in
the world. Nokia then established the Symbian Foundation,
I. T RAFFIC MONITORING AT THE ERA OF MOBILE with the intention of unifying the platform and making it
INTERNET SERVICES
open-source (Apple also partially opened its iPhone OS to
Smartphones as sensors of the built environment. software developers with the release of a software develop-
The convergence of communication and sensing on multi- ment kit). To strengthen its own mapping capabilities, Nokia
media platforms such as smartphones provides the engineer- bought Navteq, which is the largest mapping company in
ing community with unprecedented monitoring capabilities. the world, following personal navigation device manufacturer
Smartphones such as the Nokia N96 now include a video TomTom’s purchase of Tele Atlas, Navteq’s chief competitor.
camera, numerous sensors (accelerometers, light sensors, Navteq in turn owns Traffic.com, one of the leading traffic
GPS, microphone), wireless communication outlets (GSM, data collection and broadcast companies. Its competitors
GPRS, WiFi, Bluetooth, infrared), computational power and include Inrix, which provides traffic data to Microsoft’s web,
memory. This phone can be used to listen to the radio, desktop, and mobile applications.
to watch digital TV, to browse the internet, to do video
conferencing, to scan barcodes, to read pdfs, etc. The rapid New sources of traffic data for the transportation
penetration of GPS in smartphones is enabling device geopo- network. Highways have traditionally been monitored using
sitioning and context awareness, which in turn is causing static sensors, which include loop detectors built in the pave-
an explosion of Location Based Services (heavily relying ment, radars and cameras along the road, and more recently
on mapping) on the devices. For example, Nokia Maps toll tag readers (such as FasTrak or EZ-pass), which can
display theaters and museums near the phone, Google Mobile serve as probes wherever such infrastructure exists. While
provides driving directions from the phone location, and the this infrastructure has proved to be efficient for highways,
iPhone Travelocity shows hotels near the phone. Due to their the costs of deployment, communication, and maintenance
portability, computation, and communication capabilities, for such an infrastructure in the arterial network make it
smartphones are becoming useful for numerous applications prohibitive for public agencies or companies to deploy on
in which they act as sensors moving with humans embedded a global scale. To alleviate possible communication bot-
in the built infrastructure. Large scale applications include tlenecks, on October 21, 1999, the Federal Communica-
everything from population migration tracking and traffic tions Commission allocated 75MHz of spectrum as part of
flow estimation to physical activity monitoring for assisted the US Department of Transportation’s (DOT) Intelligent
living. Transportation Systems (ITS) US-wide program, with mostly
traveler safety, fuel efficiency and pollution in mind. The first
The competition for probe traffic data collection industry-government supported standard followed on August
as a proxy for the larger war to conquer the mobile 24, 2001, when ASTM’s E17.51 Standards Committee voted
internet. In recent months, there have been increased levels 20-2 to base Dedicated Short Range Communication (DSRC)
of competition between cell phone manufacturers, network on a modification of the IEEE 802.11a specification now
providers, internet service providers, computer and software named IEEE 802.11p.1 At the same time, the US DOT
manufacturers, and mapping companies. Following the tran- launched a plan which included the deployment of around
sition from desktops to laptops to smaller and more portable
devices, top companies in these industries are redefining 1 Source: Professor Raja Sengupta.
250,000 road side DSRC radios, but only led to around 100 are only available where participating vehicles / phones are
radios deployed for the entire US to this day. This example located. These are not predictable, and the local penetration
highlights the difficulty of creating a dedicated system for of devices in the network might vary. This list of problems
the transportation network. At the same time, the need of opens numerous research avenues with direct impact on
monitoring traffic remains unsolved: if traffic information technology development for traffic monitoring.
was available at a global scale including the arterial network,
several problems could potentially be solved: (1) real-time Modeling and computational challenges for monitor-
congestion estimation of the arterial network; (2) re-routing ing the transportation CPS. As indicated in the name
of the highway traffic into arterial networks where beneficial; cyberphysical, the two key components of cyberphysical
(3) optimized travel time, fuel efficient or emission optimal systems are “information” (cyber) and “constitutive laws”
routes for commuters. (modeling the physics of the system). Monitoring cyberphys-
ical systems such as the transportation network poses two
II. I MPACT OF THE MOBILE INTERNET ON THE major challenges:
TRANSPORTATION CPS
• Distributed models for the transportation network. Be-
Smartphones: a transformation from dedicated
infrastructure to market-driven technology. The scale cause GPS enabled phones sense velocity, or travel time
at which cell phones are produced, and the rate at which between two consecutive GPS readings, constitutive
they integrate new technology is dramatic. The total number models used to describe the evolution of the system
of cell phones worldwide exceeds three billion, with some need to incorporate these reading and bypass quantities
European countries with a penetration of more than 150% which cannot be measured (density, flows, counts). The
(150 cell phones for 100 people). Nokia alone produces development of such flow models, for highways and
more than 17 phones a second, which means with the arterials is still at its infancy. Techniques used for this
increasing penetration of GPS in the cellular phone fleet, include partial differential equations, queuing systems,
cell phones will soon constitute one of the major traffic and hybrid systems models of flow equations.
information sources available to the public. In North America • Machine learning models to circumvent lack of geo-
and Europe, the overwhelming majority of commuters have a graphical information. Mapping the entire US with an
cell phone, potentially populating the entire arterial network automated traffic monitoring system prevents the use
with probe traffic sensors. Obviously, the use of cellular of accurate knowledge of signage and traffic light
devices as traffic sensors has numerous benefits. (1) It is presence, let alone cycle information. The presence of
possible to leverage the market driven communication infras- stops, lights, and their effect on traffic is not available
tructure already in place. (2) The spatio-temporal penetration from databases on a US-wide scale. Furthermore, they
of cell phones in the transportation network is increasing change too often to be incorporated in flow models.
at an extremely fast pace. (3) The use of cell phones as This difficulty has to be circumvented by machine
traffic probes is device and carrier agnostic, leading to faster learning algorithms capable of learning the flow features
penetrations. (4) Major car manufacturing companies already without knowledge of the detailed infrastructure, using
have cradles and interfaces with cell phones (for example for example clustering analysis.
BMW and the iPhone) in their new cars so the sensing
information gathered by modern cars can also be sent to Spatially aware sampling and privacy. At the heart of
such monitoring system. such a system, privacy by design sampling techniques must
be used to prevent privacy invasion. In addition to proper
Lagrangian vs. Eulerian information. While cellular anonymous data collection and encryption, sampling the
phones provide an ideal bridge between the physical world vehicles at locations which are privacy safe is key to ensuring
(vehicle flows and dynamics on the road) and the information ongoing participation of the public which is needed for such
world (software systems monitoring the network), there is a system. One possible architecture for preserving privacy is
one major difference between the data collected by cell to collect data using a concept known as Virtual Trip Lines
phones and traditional data, commonly used to estimate (VTLs), which are virtual geographic line segments deployed
traffic in real time: the data collected by phones in cars across roadways in the transportation network, triggering
is Lagrangian, i.e. gathered along cars trajectories, and not phones to collect and transmit data to the system. Defining
Eulerian, i.e. control volume based. This poses major chal- optimal sampling strategies, which are privacy preserving is
lenges in building an information system for a cyberphysical still a relatively unexplored field.
infrastructure such as the transportation network. While a Real-time, online and robust availability. Unlike the
static loop detector or a camera (both Eulerian) can easily more permanent Eulerian detectors, to which data quality,
capture all vehicles going through the space monitored by reliability and performance indices can be easily attributed,
the sensor, and therefore infer from it exhaustive quantities the penetration of cell phones at a given location and time
(flows, counts, local speed), a Lagrangian sensor can only is highly variable. Furthermore, in the coming few years
monitor quantities following the vehicle, which does not give before this type of monitoring becomes the standard, the
direct access to flows, counts, etc. In addition, measurements participation of the public will be spatially and temporally
varying. This means that the algorithms used for estimating
the traffic must be robust to variability in penetration.
Inverse modeling and data assimilation. At the age
of massive data collection, one of the most fundamental
theoretical challenges associated with the reconstruction of
traffic using mobile data will be the proper use of techniques
to incorporate data into flow models or statistical models. The
development of these techniques in fields such as oceanog-
raphy or meteorology is relatively mature. For cyberphysical
systems, in particular large scale infrastructure systems, the
state of modeling, model inversion and computation is still
at its infancy and promises significant breakthroughs in the Fig. 1. Mobile Millennium traffic client. Traffic is displayed directly on
near future. the phone screen, with traffic information (construction, accidents, etc.), and
audio traffic reports. For drivers who want to opt in, the phone can also send
information to the system (Lagrangian measurements).
III. C ASE STUDY: M OBILE M ILLENNIUM
In order to study these new challenges, researchers from A back end server will aggregate data from a large
Nokia Research Center Palo Alto, Navteq, and UC Berkeley, number of mobile devices and push the data to UC Berkeley
with support from Caltrans and US DOT, have built a traffic estimation engine for data assimilation, which will combine
monitoring system using mobile devices, known as Mobile the cell phone data with other information such as loop
Millennium. Spanning 18 months, the Mobile Millennium detectors to produce the best estimate of the current state
project will implement a state-of-the-art system to collect of traffic. The map data server will provide the Navteq
traffic data from GPS-equipped mobile phones and estimate Navstreets digital map data which is required for the network
traffic conditions in real-time. The study will focus on based traffic flow models. Multiple estimation algorithms
commuters in the San Francisco Bay Area, with specific will be run in parallel as part of ongoing research, including
emphasis on monitoring the heavy congestion experienced arterial traffic models. Finally, estimate managers at UC
by travelers to and from the Lake Tahoe Ski Resorts one Berkeley and Navteq monitor the performance of the various
hundred fifty miles away. The project is a follow up of algorithms and select the best estimate to transmit back to
the Mobile Century experiment, in which 165 UC Berkeley the mobile device.
graduate students were hired to drive a 10 mile loop of
I880 in California for a day, demonstrating the feasibility B IOGRAPHIES
of a real-time traffic estimation service using GPS enabled Daniel Work is a Systems Ph.D. Student in the La-
devices only. Mobile Millennium significantly increases the grangian Sensors Systems Laboratory at the University of
scale and scope of this work by demonstrating the first real- California, Berkeley (Civil and Environmental Engineering).
time permanent monitoring system capable of using GPS His research is focused on estimation, control, and optimiza-
data from thousands of mobile devices to construct velocity tion of transportation cyberphysical systems.
fields and travel time estimates, using the VTL sampling
strategy. While the previous experiment focused on highway Alexandre Bayen received the Engineering Degree in
traffic estimation on a single segment of highway, Mobile applied mathematics from the Ecole Polytechnique, France,
Millennium will estimate traffic on all major highways in in July 1998, the M.S. degree in aeronautics and astronautics
and around the Bay Area, Sacramento, and Lake Tahoe. from Stanford University in June 1999, and the Ph.D. in
New estimation algorithms will also be implemented for aeronautics and astronautics from Stanford University in De-
determining congestion on major arterial roads which have cember 2003. He was a Visiting Researcher at NASA Ames
sufficient user penetration. Because this system will require Research Center from 2000 to 2003. Between January 2004
public participation (with an initial target of 10,000 partici- and December 2004, he worked as the Research Director of
pants), the system must be scalable, require low bandwidth, the Autonomous Navigation Laboratory at the Laboratoire de
and protect the privacy of its participants, while displaying Recherches Balistiques et Aerodynamiques, (Ministere de la
accurate estimates on a software client which should be user Defense, Vernon, France), where he holds the rank of Major.
friendly. The deployed system will be broadly characterized He has been an Assistant Professor in the Department of
by five major components: GPS-enabled smartphones in Civil and Environmental Engineering at UC Berkeley since
vehicles (driving public), a cellular network operator, cellular January 2005.
phone data aggregation, traffic estimation and traffic service
provision. On each participating mobile device, an traffic
application is executed (see Fig. 1) which is responsible
for both collecting traffic data through VTL sampling, and
displaying the current traffic estimates which are produced
from the aggregate data of all participants.